基于不变矩和神经网络的抗几何攻击水印技术

Geometric Robust Watermarking Through Moment Invariants and Neural Network

针对数字水印领域中抗几何攻击这一难题,提出了一种新颖的盲水印方法。该方法充分利用BP网络具有很强的非线性逼近能力这一特点,建立含水印图象不变矩与水印不变矩之间的对应关系,因而在检测水印时不需原始图象。在检测水印时,提出了一种利用不变矩特征距离来计算水印相关性的方法。不变矩和神经网络的结合,不仅可以提高水印的强壮性,使水印算法具有抵抗几何攻击的能力,还可以充分利用神经网络的强壮性使水印的检测更加可靠。实验表明该方法具有良好的效果,不仅可以抵抗几何攻击,而且在高压缩比的JPEG压缩和一些图象处理下仍可检测到水印。同时该方法具有可推广性,可以进一步方便地应用于其它类型和不同变换域中的数字水印方法。

Aimed at the problem of resisting to geometric attacks in digital watermarking field, a novel blind watermark method was presented. The method took full advantages of the nonlinear mapping characteristics of the BP network, set up the moment invariants relationship between the watermarked image and the watermark, and the original image was not required to detect the watermark. When detecting the watermark, the similarity of the watermarks was calculated using the eigen distance of the moment invariants. The combination of moment invariants and neural network can improve robustness to geometric attacks and make the watermark detection more reliable. The experimental results showed that the method is effective. It is robust not only to geometric attacks, but also to some image processing operations including JPEG compression with high ratio. Furthermore, the method can be extended to some other digital watermark methods in different transform domains.